Cor.inc Blog
Clear notes from AI, product, and engineering work, written for readers who need the practical decision points first.
On October 29, 2026, at an internal corporate study session led by Representative Kousuke Terada, we tested two techniques on a small language model (Qwen2.5-3B-Instruct): "providing examples" and "asking for a review." Based on questions that improved, worsened, or confidently asserted an error was correct, we explain how to evaluate generative AI for business operations.
Before using a local inference API inside the company, design authentication, an allow list and device retirement as separate things. An allow list shows which peer may be reached. Who may use it for which work is decided separately.
Field AI that runs near cameras and sensors, and a local LLM that handles internal documents, care about different conditions. Cor.Inc.'s policy is to separate the verification environments. Separation itself is not proof of safety.
This is the story of how our representative, Kousuke Terada—a former trumpet player diagnosed with ASD at age 28—founded the AI company Cor.Inc. after working in sales at a musical instrument manufacturer and Recruit. He writes about why he is using technology to give shape to what words alone cannot convey.
This article separates what can be counted from a git history under stated rules, and what cannot. Four real OSS projects were checked at repository level. It does not guarantee a person's ability or that the record is complete.
Grift CLI promises not to output scores or rankings. During a cross-check in October 2026, it was discovered that while notes stated it was not a ranking, the table was sorted in descending order of commit count. We plainly explain the five acceptance criteria imposed on version 0.8 for clients, SES sales representatives, and engineers.
This article explains why contract development estimates vary from person to person, and how to keep an explainable basis by recording the breakdown and the assumptions. The estimation-support web app ended in September 2026. Current consultations are for AI contract development and adoption support.
For executives and administrative departments unsure whether they can input contracts, meeting minutes, estimates, or customer information into generative AI, this article explains how to classify confidentiality levels, choose between cloud AI and local LLMs, and establish internal rules.
Given the current 20.4% AI adoption rate among SMEs, this article explains how to start utilizing AI without overextension, beginning with general affairs and administrative departments. Written for companies considering their first steps in AI utilization.
For executives and administrative departments formulating rules for generative AI use, this article outlines prohibited input information, approved AI tools, logging, education, and incident response. It is intended for companies seeking to balance AI utilization with information management.
We explain why internal document search and RAG implementations fail to deliver results, looking from the perspectives of data organization, permissions, evaluation metrics, and operational design. This is a practical guide to keep internal knowledge AI from ending at the PoC stage.
We explain the basics of local LLMs, how they differ from cloud AI, the types of operations best suited for adoption, and the security and operational requirements to verify before a PoC. This article is for companies looking to securely leverage AI with confidential data.
